检索结果-内蒙古大学图书馆

Effective speech emotion recognition using deep learning approaches for Algerian dialect

学校读者我要写书评

暂无评论

Effective speech emotion recognition using deep learning app...

2021 International Conference of Women in Data Science at Taif University, WiDSTaif 2021

作者： Yahia Cherif, Raoudha Moussaoui, Abdelouahab Frahta, Nabila Berrimi, Mohamed Dep. of Computer Science University of Ferhat Abbes Setif I Setif Algeria Dep. of Technology University of Ferhat Abbes Setif I Setif Algeria

ISBN: (纸本)9781665449489

Speech Emotion Recognition is an active research problem in the field of deep learning. Where the goal is to detect different emotions within audio speech using deep artificial neural networks. In this work, we focus on speech emotion recognition in the Algerian dialect. The paper introduces a new large Algerian speech emotion dataset collected from different Algerian TV shows. The audio signals were manually annotated as happy, angry, neutral, or sad. After the data collection, we applied several classification methods such as machine learning-based models, convolutional neural networks (CNNs), Long Short Term Memory (LSTM) networks, and Bidirectional LSTM (BLSTM). After applying the proposed models on our dataset, the best classification accuracy is 93.34% achieved by the LSTM-CNN model. © 2021 IEEE.

关键词： Long short-term memory

Energy Access Assessment Report Focusing on Energy Poverty and Energy Demand for Rwanda

学校读者我要写书评

暂无评论

Energy Access Assessment Report Focusing on Energy Poverty a...

Sciences and Techniques of Automatic Control and computer Engineering MI-STA, IEEE International Maghreb Meeting of the Conference on

作者： Adimchinobi Asiegbu Ali Almaktoof Abdurazag Khalat Dep. of Electrical Electronics and Computer Engineering Cape Peninsula University of Technology Cape Town South Africa Dep. of Electrical and Electronics Engineering Sabratha University Sabratha Libya

This technical report is a researched document that examines energy access in Rwanda, focusing on energy poverty assessment and energy demand assessment. The key findings and observations can help mitigate the energy access problems and improve access to modern energy in Rwanda. Studies have shown that sub-Saharan African countries have poor access or no access to modern energy, Rwanda being part of sub-Saharan Africa also has Poor access to modern and clean energy. To carefully study these energy access problems and proffer solutions to increase energy access in Rwanda based on realities in today’s Africa, (where most households fall below poverty lines, with social issues, low Purchasing Power Parity (PPP) and Gross Domestic Power (GDP)), an assessment of these energy problems is done based on Energy Poverty Assessment (EPA) and Energy Demand Assessment (EDA) Composite Multidimensional Indicators (CMDI), respectively. Based on the findings of this research, non-poor energy households increased their energy consumption by 29.26% to 0.68 billion kWh. The clean cooking gas usage in Rwanda is very low at 2.1%, The 2016 estimated electricity consumption is 0.5273 billion kWh, growing by 57.41% to 0.83 billion kWh for energy-poor households in 2021, compared to 52.6% predicted by International Energy Agency (IEA). Finally, there is a direct relationship between energy access and economic development including quality of life, standard of living, per capita income, urbanization, modern infrastructure, etc. Thus, the key findings and observations in this research report will improve economic and infrastructural development in Rwanda, if implemented.

关键词：

Logistic Regression Classification for Assessing the Risk of Kidney Tumor

学校读者我要写书评

暂无评论

Logistic Regression Classification for Assessing the Risk of...

Electrical, Energy, and Artificial Intelligence (EICEEAI), International Engineering Conference on

作者： Dalia Alzu'bi Rabia Emhamed Al Mamlook Ahmad Nasayreh Mohammad Aljaidi Rami AlAzab Hasan Gharaibeh Qais Al-Na'amneh Dep. Information and Systems Engineering Concordia University Montreal Canada Dep. of Business Administration Trine University Indiana USA Department of Industrial Engineering University Zawia Tripoli Libia Dep. Computer Science Yarmouk University Irbid Jordan Dep. of Computer Science Zarqa University Zarqa Jordan Dep. Surgery and Urology Jordan University of Science and Technology Irbid Jordan Dep. of Cyber Security Applied Science Private University Amman Jordan

ISBN: (数字)9798350373363

ISBN: (纸本)9798350373370

The kidneys have the important function of filtering waste products and toxins from the blood in the human body. Any problems affecting the kidneys can potentially impact their efficiency and overall function. Kidney Tumor (KT) is a disease that specifically affects kidney cells and causes the abnormal growth of tissue in one or both kidneys. Assessing the risk of developing a kidney tumor and identifying the key factors that contribute to this risk are crucial for ensuring patient well-being. Additionally, this information assists doctors and specialists in making faster and more accurate diagnoses, determining appropriate treatment methods, and potentially influencing the course of the disease to minimize its severity and impact. Machine learning algorithms have recently been introduced to evaluate disease risks, and in this study, we specifically focus on examining the risk of kidney tumor development and investigating the influencing factors. We employed a logistic regression model to predict if a patient is at risk of developing a kidney tumor or not, and further categorized the samples into high-risk and low-risk groups. Our model was trained and tested using a unique dataset obtained from the (KAUH) hospital in Jordan, consisting of well-balanced metadata from 120 patients with kidney issues. Our work demonstrated accuracy results ranging from 90% to 98%. Ultimately, specialists can utilize our model as an additional tool to enhance the speed and accuracy of patient diagnoses

关键词： Logistic regression Analytical models Accuracy Predictive models Genetics History Reliability

BiRT: Bio-inspired Replay in Vision Transformers for Continual Learning

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Jeeveswaran, Kishaan Bhat, Prashant Zonooz, Bahram Arani, Elahe Advanced Research Lab NavInfo Europe Netherlands Dep. of Mathematics and Computer Science Eindhoven University of Technology Netherlands

The ability of deep neural networks to continually learn and adapt to a sequence of tasks has remained challenging due to catastrophic forgetting of previously learned tasks. Humans, on the other hand, have a remarkable ability to acquire, assimilate, and transfer knowledge across tasks throughout their lifetime without catastrophic forgetting. The versatility of the brain can be attributed to the rehearsal of abstract experiences through a complementary learning system. However, representation rehearsal in vision transformers lacks diversity, resulting in overfitting and consequently, performance drops significantly compared to raw image rehearsal. Therefore, we propose BiRT, a novel representation rehearsal-based continual learning approach using vision transformers. Specifically, we introduce constructive noises at various stages of the vision transformer and enforce consistency in predictions with respect to an exponential moving average of the working model. Our method provides consistent performance gain over raw image and vanilla representation rehearsal on several challenging CL benchmarks, while being memory efficient and robust to natural and adversarial corruptions. 1 © 2023, CC BY-NC-ND.

关键词： Deep neural networks

Multi-Task Structural Learning using Local Task Similarity induced Neuron Creation and Removal

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Gurulingan, Naresh Kumar Zonooz, Bahram Arani, Elahe Advanced Research Lab NavInfo Europe Netherlands Dep. of Mathematics and Computer Science Eindhoven University of Technology Netherlands

Multi-task learning has the potential to improve generalization by maximizing positive transfer between tasks while reducing task interference. Fully achieving this potential is hindered by manually designed architectures that remain static throughout training. On the contrary, learning in the brain occurs through structural changes that are in tandem with changes in synaptic strength. Thus, we propose Multi-Task Structural Learning (MTSL) that simultaneously learns the multi-task architecture and its parameters. MTSL begins with an identical single-task network for each task and alternates between a task-learning phase and a structural-learning phase. In the task learning phase, each network specializes in the corresponding task. In each of the structural learning phases, starting from the earliest layer, locally similar task layers first transfer their knowledge to a newly created group layer before being removed. MTSL then uses the group layer in place of the corresponding removed task layers and moves on to the next layers. Our empirical results show that MTSL achieves competitive generalization with various baselines and improves robustness to out-of-distribution data. Code available at https://***/NeurAI-Lab/MTSL. © 2023, CC BY-NC-ND.

关键词： Network architecture

Brain Tumor Classification System Using Decision Tree Algorithm

学校读者我要写书评

暂无评论

Brain Tumor Classification System Using Decision Tree Algori...

2025 IEEE International Conference on Cybernetics and Innovations, ICCI 2025

作者： Thaiparnit, Sattarpoom Janchidfah, Thanakrit Farooq, Adil Rajamangala University of Technology Suvarnabhumi Faculty of Business Administration Dep. Of Information System and Computer Business Suphan Buri72130 Thailand Muban Chombueng Rajabhat University dep. Of Digital Media Technology Ratchaburi70150 Thailand Maynooth University Department of Electronic Engineering Ireland

ISBN: (纸本)9798331533267

This research aims to study, design, and develop a brain tumor classification system using artificial intelligence, specifically decision tree algorithms. The system's primary objective is to assist medical personnel in making decisions, as distinguishing between two types of brain tumors-benign and malignant (cancerous)-can be challenging, especially for less experienced medical staff. Due to the similarities between these two types of tumors, it may be difficult for medical practitioners with limited expertise to differentiate between them in a timely manner. The proposed brain tumor classification system is designed to address this challenge. The research team conducted the study over approximately seven months, starting from the second semester of the 2023 academic year through the first semester of the 2024 academic year. Experimental results revealed that among the tree-based algorithms, Random Forest achieved the highest accuracy, followed by M5P. The best average accuracy from the experiments was obtained using a 70% training and 30% testing split, yielding an accuracy of 96.83%. Therefore, the research team selected the 70% training and 30% testing split for model development, as it provided the highest accuracy compared to the 60% training and 40% testing split, and the 80% training and 20% testing split, which showed lower accuracy. The results demonstrate that the 70% training and 30% testing approach is optimal for the brain tumor classification system, ensuring high accuracy in classifying brain tumors. © 2025 IEEE.

关键词： Random forests

TASK-AWARE INFORMATION ROUTING FROM COMMON REPRESENTATION SPACE IN LIFELONG LEARNING

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Bhat, Prashant Zonooz, Bahram Arani, Elahe Advanced Research Lab NavInfo Europe Netherlands Dep. of Mathematics and Computer Science Eindhoven University of Technology Netherlands

Intelligent systems dep.oyed in the real world suffer from catastrophic forgetting when exposed to a sequence of tasks. Humans, on the other hand, acquire, consolidate, and transfer knowledge between tasks that rarely interfere with the consolidated knowledge. Accompanied by self-regulated neurogenesis, continual learning in the brain is governed by a rich set of neurophysiological processes that harbor different types of knowledge, which are then integrated by conscious processing. Thus, inspired by the Global Workspace Theory of conscious information access in the brain, we propose TAMiL, a continual learning method that entails task-attention modules to capture task-specific information from the common representation space. We employ simple, undercomplete autoencoders to create a communication bottleneck between the common representation space and the global workspace, allowing only the task-relevant information to the global workspace, thus greatly reducing task interference. Experimental results show that our method outperforms state-of-the-art rehearsal-based and dynamic sparse approaches and bridges the gap between fixed capacity and parameter isolation approaches while being scalable. We also show that our method effectively mitigates catastrophic forgetting while being well-calibrated with reduced task-recency bias. © 2023, CC BY-NC-ND.

关键词： Intelligent systems

ERROR SENSITIVITY MODULATION BASED EXPERIENCE REPLAY: MITIGATING ABRUPT REPRESENTATION DRIFT IN CONTINUAL LEARNING

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Sarfraz, Fahad Arani, Elahe Zonooz, Bahram Advanced Research Lab NavInfo Europe Netherlands Dep. of Mathematics and Computer Science Eindhoven University of Technology Netherlands

Humans excel at lifelong learning, as the brain has evolved to be robust to distribution shifts and noise in our ever-changing environment. Deep neural networks (DNNs), however, exhibit catastrophic forgetting and the learned representations drift drastically as they encounter a new task. This alludes to a different error-based learning mechanism in the brain. Unlike DNNs, where learning scales linearly with the magnitude of the error, the sensitivity to errors in the brain decreases as a function of their magnitude. To this end, we propose ESMER which employs a principled mechanism to modulate error sensitivity in a dual-memory rehearsal-based system. Concretely, it maintains a memory of past errors and uses it to modify the learning dynamics so that the model learns more from small consistent errors compared to large sudden errors. We also propose Error-Sensitive Reservoir Sampling to maintain episodic memory, which leverages the error history to pre-select low-loss samples as candidates for the buffer, which are better suited for retaining information. Empirical results show that ESMER effectively reduces forgetting and abrupt drift in representations at the task boundary by gradually adapting to the new task while consolidating knowledge. Remarkably, it also enables the model to learn under high levels of label noise, which is ubiquitous in real-world data streams. Code: https://***/NeurAI-Lab/ESMER. © 2023, CC BY-NC-ND.

关键词： Errors

Poster: Selection of Optimal Neural Model using Spiking Neural Network for Edge Computing*

学校读者我要写书评

暂无评论

Poster: Selection of Optimal Neural Model using Spiking Neur...

International Conference on Distributed Computing Systems

作者： Sanaullah Kaushik Roy Ulrich Ruckert Thorsten Jungeblut Dep. of Eng. and Math. Industrial IOT University of Applied Sciences and Arts Bielefeld Germany Dep. of Computer Science Cyber Defense and AI Lab Lab North Carolina A&T State University NC USA Dep. of Faculty of Technology Cognitive Interaction Technology Universitat Bielefeld Bielefeld Germany

ISBN: (数字)9798350386059

ISBN: (纸本)9798350386066

Spiking Neural Networks (SNNs), are inspired by the biological brain's complicated signaling mechanisms and possess unique characteristics that set them apart from traditional artificial neural networks. This research study explores the challenging domain of image classification, specifically utilizing the well-known MNIST dataset through the development and thorough evaluation of different neural models for edge computing. However, the primary contribution is the autonomous selection of the best-performing SNN model through various early stopping approaches and validation functions, allowing the models to autonomously adapt during training. In addition, this article presents the standalone AutoML-SNN model, which is the introduction of dynamic elements into selected SNN domains, enhancing their adaptability to complex patterns within the dataset. Furthermore, the early stopping methodologies are used to reduce overfitting hazards, and using the 3000-neuron set, the LIF appeared as the most proficient neural model.

关键词： Training Adaptation models Computational modeling Biological system modeling Neurons Spiking neural networks Hazards